Asset Allocation Approaches: Building From the Bottom Up
by James B. Cloonan
In my column last month, I argued that the popular approach to asset allocation was seriously flawed because it treated asset allocation as an end, rather than as a means to an end. The real objective is risk reduction, and asset allocation is a way to get there.
The current popular asset allocation approaches divide investments into asset classes—for example, stocks, bonds and real estate. The approach then divides your investible assets between those asset classes based on some formula, which is often related to an individuals age or life cycle. These approaches also suggest you further divide up your investment dollars among the various sub-classes—for instance, small-cap and large-cap stocks, U.S. and foreign stocks, etc. The next step is allocation among individual stocks. This is normally called diversification, but it is really the same concept as asset allocation.
My criticism is that the popular asset allocation approaches leave out an extremely important step. The ultimate goal is to reduce risk, and there is no question that allocation across asset classes can reduce risk. However, it cant reduce risk effectively unless you know which investments in each class are involved. As I pointed out, some individual bonds are riskier than the average stock, and certain bond portfolios are riskier than certain stock portfolios. The important issue is how the holdings correlate. What an individual needs to do to reduce risk is to add to his portfolio those holdings that have both good prospects and low correlations with the other holdings. Whether those assets belong to a particular class is not important.
I believe a more effective approach to controlling risk is to start from the bottom and find investments with the potential for good returns, and then to choose a portfolio from among those investments. The objective, once again, is to get the highest return within an acceptable level of risk. To control risk (unless you have a very high risk tolerance), it will probably be necessary to have some assets that are not equities or are a very different form of equity, such as real estate investment trusts (REITS) or gold-mining stocks.
The Popular Approach
Heres how the popular approach typically works.
Table 1 presents a list I prepared of major asset classes and sub-classes (using index funds and exchange-traded index funds [ETFs] as surrogates), with their long-term (10-year) average returns and their shorter term (six-month) average risk level.
The 10-year time period was chosen because most individuals have longer-term holding periods, and I feel the 10-year returns are the best estimate of long-term future returns.
I used a shorter period for risk measurement because short-term volatility reduces long-term returns. It is the change in the risk levels of asset classes that should lead you to adjust your allocation, and not the maintenance of some arbitrary percentage—but Ill get into that later in this column.
The RiskGrade index is used as the measure of risk. The RiskGrade is drawn from the RiskGrade Web site, which provides risk analysis of portfolios and securities. RiskGrades are based on standard deviation, the amount by which most actual returns over a given period vary around the average return, but are standardized relative to all the worlds equities. A RiskGrade of 77 implies a risk that is 77% as high as the average risk of all equities.
In Table 1, I also show what percentage of each asset class could be combined with Treasury bills to achieve a certain level of risk, in this instance a RiskGrade of 61—which is the current risk level of the widely touted allocation 60% stocks/40% bonds. For example, a portfolio consisting of 55% large-cap stocks and 45% Treasury bills would achieve a RiskGrade of 61.
Table 2 looks at some sample asset allocations and provides each portfolios RiskGrade and expected return (based on the long-term historical return). The first two examples are typical recommendations among the major asset classes, and the second two examples are typical recommendations that include subcategories.
How do these portfolios stack up?
In the last issue, I suggested that individuals use a portfolio benchmark that consists of the highest-returning asset class combined with Treasury bills to reduce the risk to your appropriate risk level. The risk level in our examples here is 61, and the highest-returning asset class has changed slightly from a month ago and is now small-cap value. [I believe, based on shorter-term results, that a micro-cap value would beat a small-cap value portfolio, but we do not have a perfect surrogate for micro-cap value.]
The benchmark combination that achieves the desired RiskGrade of 61 is 60% small-cap value and 40% Treasury bills, with a return of 10.69%—thats the benchmark. In this example, we should not consider allocations with poorer results than this benchmark.
None of the suggested allocations in Table 2 match this benchmark. In fact, in putting this data together, it was easy to see that allocations alone across the various sub-classes of equities did not provide much risk reduction.
|TABLE 2. Sample Asset Allocations (based on data as of 8/30/02)|
|Portfolio 1: 60% Stocks/40% Bonds||61***||8.73|
|Portfolio 2: 80% Stocks/20% Bonds||83||9.27|
|Portfolio 3: see below||76||9.23|
|Portfolio 4: see below||70||8.62|
|Allocations||Portfolio 3||Portfolio 4|
* Surrogates used in this table are the same as those used in Table 1; for Portfolios 1 and 2, stocks are all U.S. stocks (represented by VTSMX) and bonds are all bonds (represented by VBMFX).
** Based on 10-year historical average annual return.
*** RiskGrade of 61 used for normalization in Table 1.
Building from the Bottom
I believe a much more productive approach to risk reduction is to choose a portfolio of stocks with low correlations, regardless of asset class. You can use the RiskGrade Web site to develop such a portfolio—the RiskGrade of a portfolio takes into consideration correlation, since a portfolio of low correlation stocks will have lower volatility and therefore a lower RiskGrade.
The Beginners Portfolio diversifies away over 60% of the individual stock risk. [For an explanation of the experimental Beginners Portfolio, see The AAII Beginners Portfolio: An Annual Performance Review, August 2001 AAII Journal.] In our new service, the Stock Superstars Report, the final stock choice is based on which new recommended stock helps reduce portfolio risk the most.
The returns and risk levels in Table 1 and Table 2 are averages for asset classes. Well-chosen portfolios—portfolios in which the individual securities that are selected have good return prospects and low return correlations—will have much better return/risk ratios than portfolios with securities that are the average in each asset class.
In addition, the indexes and mutual funds used consist of portfolios that are capitalization weighted—a greater percentage of the portfolio is invested in companies with larger market capitalizations. Since smaller-cap stocks outperform larger-cap stocks, the same index stocks and portfolios would perform better if equally weighted. Ill pursue this issue again in future articles.
What about changing allocations over time?
The traditional approach would be that if the accepted allocation was 60% stocks and 40% bonds and the stock market ran up, there would be the need to sell stocks and buy bonds to maintain the original percentage allocation.
The problem here is once again that the approach is losing sight of the true objective—to control risk. Perhaps selling stocks and buying more bonds when the market goes up is a good idea sometimes.
But, maybe not. Imagine that the U.S. moves into a bull market. The first year would likely be very strong, and the conventional wisdom to readjust the stock/bond mix would become necessary.
But what if—and it is not unlikely—because stocks are moving up with some consistency, their risk is going down, and with the economy starting to perk up and the Fed doing its thing, bonds are becoming more volatile (risky).
The alternative rebalancing approach that I suggest is to maintain your given risk level. And to maintain the risk level you have chosen, you might find you should actually be buying stocks and selling bonds as stocks go up. With RiskGrades or other sources of volatility information, you can monitor the risk of your portfolio on a continual basis and make adjustments as needed.
One word of caution. Regardless of the approach to rebalancing, dont do it too often. Annually is more than enough. Not only are there transaction costs and perhaps taxes involved, but there is strong evidence of some serial correlation—that means assets going up tend to keep going up.
The Bottom Line
Asset allocation—also known as diversification—is an extremely important concept for reducing risk. The question is: How do you get there?
Artificial percentages that try to achieve diversification are primitive and outdated in the computerized age when we can deal with other more direct approaches.
I am not suggesting that you focus solely on a particular RiskGrade number. For instance, theoretically, you could end up with a portfolio consisting of 20% tech stocks and 80% T-bills. But I wouldnt recommend it—risk can change rapidly, and for that reason it probably isnt wise to be in a single asset sub-class. Unless you are monitoring the risk of your portfolio on a regular basis, it is wise to be more conservative and look at longer-term volatility. Remember, too, that the numbers you are dealing with are based on historical returns, and those returns and correlations wont necessarily be repeated.
What I am suggesting is that you can achieve diversification by focusing on individual selection and correlations among those selections that can be measured directly, rather than using categories as a proxy.
In the past two columns, I have discussed the meaning of asset allocation and diversification and how they can be used for risk reduction. I have pointed out the weaknesses of the current approaches and, in particular, the problem with accepting ones that are too simplistic.
While I have alluded to some better approaches, I have not yet provided specific solutions to the problems discussed. Ill do that in the next column.
James B. Cloonan is chairman of AAII.